[net.micro.68k] 68000 vs 6502

collinge@uvicctr.UUCP (Doug Collinge) (09/13/85)

I have an application which is currently running on a 6502.
I am thinking of replacing four 6502 machines with an Atari ST.
One of the tasks of the CPU is to transfer 1024 bytes from
an I/O device into RAM.  I was wondering if the 8MHz 68k in the
Atari could do four such transfers in the time it took a 1MHz
6502 to do it.  Having computed the cycles I find that, yes,
the 68k can just barely do it even though its memory bandwidth
is twice that of the 6502.  So for this application the 68k
is the same speed as a 4MHz 6502!  I think this is very funny.
I salute the designers of the 6502 for making such a remarkably
durable little machine.

Yes, I understand that this is what the 6502 does best and also
what the 68k does worst.  And other factors make the 68k desirable
here so I will probably use it anyway.

I understand that the 68010 has a little cache in it to speed up
loops just like this one.  Can we drop a 68010 into an ST and
expect it to work?  Has anyone tried it yet?
-- 
		Doug Collinge
		School of Music, University of Victoria,
		PO Box 1700, Victoria, B.C.,
		Canada,  V8W 2Y2  
		decvax!nrl-css!uvicctr!collinge
		decvax!uw-beaver!uvicctr!collinge
		ubc-vision!uvicctr!collinge

david@prism.UUCP (09/20/85)

	Why not drop in 65802's to replace your 6502's?

	The 65802 is the 6502 pin-compatible version of the
	65816, the new 16-bit 6502.  The 65816 features 16 
	megabyte adressing, user registers that can be expanded
	from eight bits to sixteen, a sixteen-bit stack pointer,
	a relocatable zero page (now called the direct page), a block
	move instruction, and much more.  In its emulation mode,
	it emulates a 6502 down to the weird invalid addresses
	needed to operate the Apple disk II controller; in native
	mode, you already know over half the opcodes if you know how to
	program a 6502.

	If a 1MHz 6502 can keep up with an 8MHz 68000, what will happen
	when somebody builds a machine using a 4 or 5 MHz 65816 --
	which will be exactly four or five times faster, since the
	memory access cycle is the same as the clock cycle?  (The
	standard 65802/65816 part is 4MHz, and it's CMOS).

	For more information, contact the Western Design Center, 2166
	East Brown Road, Mesa, AZ, 602-962-4545 (it is also being
	sourced by GTE Microcircuits).

		David Eyes
		Mirror Systems, Inc.

		{cca, ihnp4, inmet, mit-eddie, wjh12} ...
				... mirror!prism!david

collinge@uvicctr.UUCP (Doug Collinge) (09/28/85)

I must confess that I made something of an error in the earlier posting.
I didn't know about the DBcc instruction, which makes things go a little
faster for the 68k.  Not that much - about 50%.

With respect to the 65802: has anyone actually bought one of these things
and plugged it into their Apple or whatever?  If so please post.  What
do they currently cost?  It was $90US last time I called.

-- 
		Doug Collinge
		School of Music, University of Victoria,
		PO Box 1700, Victoria, B.C.,
		Canada,  V8W 2Y2  
		decvax!nrl-css!uvicctr!collinge
		decvax!uw-beaver!uvicctr!collinge
		ubc-vision!uvicctr!collinge

rbt@sftig.UUCP (R.Thomas) (10/02/85)

> 
> 
> 
> 
> 	Why not drop in 65802's to replace your 6502's?
> 
> 	The 65802 is the 6502 pin-compatible version of the
> 	65816, the new 16-bit 6502.  The 65816 features 16 
> 	megabyte adressing, user registers that can be expanded
> 	from eight bits to sixteen, a sixteen-bit stack pointer,
> 	a relocatable zero page (now called the direct page), a block
> 	move instruction, and much more.  In its emulation mode,
> 	it emulates a 6502 down to the weird invalid addresses
> 	needed to operate the Apple disk II controller; in native
> 	mode, you already know over half the opcodes if you know how to
> 	program a 6502.
> 
> 	If a 1MHz 6502 can keep up with an 8MHz 68000, what will happen
> 	when somebody builds a machine using a 4 or 5 MHz 65816 --
> 	which will be exactly four or five times faster, since the
> 	memory access cycle is the same as the clock cycle?  (The
> 	standard 65802/65816 part is 4MHz, and it's CMOS).
> 
> 	For more information, contact the Western Design Center, 2166
> 	East Brown Road, Mesa, AZ, 602-962-4545 (it is also being
> 	sourced by GTE Microcircuits).
> 
> 		David Eyes
> 		Mirror Systems, Inc.
> 
> 		{cca, ihnp4, inmet, mit-eddie, wjh12} ...
> 				... mirror!prism!david

Does anybody know whether you can replace the 65(C)02 in an Apple IIe/+/c with
the 65802 without causing serious compatibility problems?  If you do, and if
it works, what do you get for your efforts?  No chance of direct
addressability to memory beyond the II's hacked over 64K, of course, but do
the 16 bit registers work?, howabout the 'direct' page stuff?, the block move
instructions?  Does it mess up any of the timing loops (like in the disk driver
software, and the paddle reader software)?  Or did they preserve the same
number of clocks per instruction as the 6502 as a compatibility 'feature'?

Interesting,

RIck Thomas
ihnp4!attunix!rbt

henry@utzoo.UUCP (Henry Spencer) (10/02/85)

> 	Why not drop in 65802's to replace your 6502's?
> 
> 	The 65802 is the 6502 pin-compatible version of the
> 	65816, the new 16-bit 6502.  The 65816 features 16 
> 	megabyte adressing...

As I recall, the 65802 does big-memory addressing the same way the 8086
does, i.e. by what is essentially on-chip bank switching rather than by
actually being able to generate and manipulate 24-bit addresses as
single entities.  Yuck.  I wouldn't touch it with a ten-foot pole.
-- 
				Henry Spencer @ U of Toronto Zoology
				{allegra,ihnp4,linus,decvax}!utzoo!henry

collinge@uvicctr.UUCP (Doug Collinge) (10/06/85)

In article <6012@utzoo.UUCP> henry@utzoo.UUCP (Henry Spencer) writes:

>As I recall, the 65802 does big-memory addressing the same way the 8086
>does, i.e. by what is essentially on-chip bank switching rather than by
>actually being able to generate and manipulate 24-bit addresses as
>single entities.  
Yes, this is true.  The 65802 is to the 6502 as the 8088 is to the 8080.

>Yuck.  I wouldn't touch it with a ten-foot pole.
Yes - you certainly wouldn't want to design it into anything.  If it
drops into Apples, though, I might buy one - just because it would be
easier to program.  Note that a 65802 can't address more than 64K
anyway so the yukkiness of the bank-switching is immaterial.  The
65816 is the counterpart to the 8086.
-- 
		Doug Collinge
		School of Music, University of Victoria,
		PO Box 1700, Victoria, B.C.,
		Canada,  V8W 2Y2  
		decvax!nrl-css!uvicctr!collinge
		decvax!uw-beaver!uvicctr!collinge
		ubc-vision!uvicctr!collinge

ix469@sdcc6.UUCP (david smith) (10/06/85)

The 65816 is indeed pin-compatible with the 6502 and it may be
possible to "drop" one into an Apple II without major problems,
provided you don't want to address more than 64K.  In order to
address the 16 megabytes, Western Design uses the data bus as
address bus for the half CPU cycle that the data bus is not used.
Unfortunately, during the time, the video is busy playing with
things.  I believe that the MMU switches the CPU off the data bus
during that time.


==================
David L. Smith
UC Sandy Eggo
{ucbvax, ihnp4}!sdcsvax!sdcc6!ix469

All opinions expressed herein are strictly the products of my
automated posting facility and have no affiliation with myself, my
employers or my cat.

ralphw@ius2.cs.cmu.edu.ARPA (Ralph Hyre) (10/17/85)

In article <2230@sdcc6.UUCP> ix469@sdcc6.UUCP (david smith{|ix}) writes:
>The 65816 is indeed pin-compatible with the 6502 and it may be
>possible to "drop" one into an Apple II without major problems,
>provided you don't want to address more than 64K.

I thought only the 65802 was pin-compatible.  Has anyone
actually made this substitution?

>In order to address the 16 megabytes, Western Design uses the data
>bus as address bus for the half CPU cycle that the data bus is not
>used.  Unfortunately, during the time, the video is busy playing with
>things.

Apple // display circuitry should be redone anyway, they should
give up on this color in even or odd column stuff.  As long as
they keep the entry points for text and graphics routines the same,
most stuff should run OK.

>I believe that the MMU switches the CPU off the data bus during that time.

This seems to imply that Apple was thinking about bypassing the MMU
at some later point to get at a large bank of memory.  Does the recently
accounced Apple memory card have a linear-addressing coprocessor port that
would allow a 65816 to address more than 64K without bank switching?
(The Applied Engineering Ramworks card does, is AE planning to support people
putting in 65816's in their machines by putting an extender socket which would
connect the data bus to the coprocessor port before it gets to the MMU chip on
the //e motherboard?)

If all of this is so, will we see 65816-enhanced Apple // come January? 
(All the sales people tell me is that the next version of the //e
will be the same shade of white as the //c and the Unidisk drive.)
--
				- Ralph
Internet: ralphw@c.cs.cmu.edu (cmu-cs-c.arpa)
Usenet: ralphw@mit-eddie.uucp
Fidonet: Ralph Hyre at Fido #385 Pitt-Bull (or maybe Net 129, node 0)
Snail Mail: don't bother
-- 
				- Ralph
Internet: ralphw@c.cs.cmu.edu (cmu-cs-c.arpa)
Usenet: ralphw@mit-eddie.uucp
Fidonet: Ralph Hyre at Fido #385 Pitt-Bull (or maybe Net 129, node 0)
Snail Mail: don't bother

ix469@sdcc6.UUCP (david smith) (10/22/85)

In article <216@ius2.cs.cmu.edu.ARPA> ralphw@ius2.cs.cmu.edu.ARPA (Ralph Hyre) writes:
>In article <2230@sdcc6.UUCP> ix469@sdcc6.UUCP (david smith{|ix}) writes:
>>The 65816 is indeed pin-compatible with the 6502 and it may be
>>possible to "drop" one into an Apple II without major problems,
>>provided you don't want to address more than 64K.
>
>I thought only the 65802 was pin-compatible.  Has anyone
>actually made this substitution?
>
According to my spec sheet, it is indeed pin-compatible.  The data
bus doubles as the high eight bits of the address bus during the
time that a "normal:-)" 6502 wouldn't be using it.

>>Unfortunately, during the time, the video is busy playing with
>>things.
>
>Apple // display circuitry should be redone anyway, they should
>give up on this color in even or odd column stuff.  As long as
>they keep the entry points for text and graphics routines the same,
>most stuff should run OK.
If you redo the video circuitry, you may as well trash your Apple,
since all you're using at that point is the keyboard, slots and
power supply.  I was thinking of putting a 65816 on a board and
running it at 4Mhz, or faster if possible and using the 6502 as an
i/o processor.  I would like to be able to let the 65816 handle the
video display, because it's got such neat tricks built in, like move
a block of memory, etc, but I can't think of a neat way to do this without
slowing it down to 1Mhz.  Perhaps a wait state while the 65816 was
addressing low memory would do the trick (slowing it to only 2Mhz).  
The disk drives would have to be handled by the 6502 (it would be 
preferable, too!).

==================
David L. Smith
UC Sandy Eggo
{ucbvax, ihnp4}!sdcsvax!sdcc6!ix469

All opinions expressed herein are strictly the products of my
automated posting facility and have no affiliation with myself, my
employers or my cat.

rbt@sftig.UUCP (R.Thomas) (11/04/85)

>                I was thinking of putting a 65816 on a board and
> running it at 4Mhz, or faster if possible and using the 6502 as an
> i/o processor.  I would like to be able to let the 65816 handle the
> video display, because it's got such neat tricks built in, like move
> a block of memory, etc, but I can't think of a neat way to do this without
> slowing it down to 1Mhz.  Perhaps a wait state while the 65816 was
> addressing low memory would do the trick (slowing it to only 2Mhz).  
> The disk drives would have to be handled by the 6502 (it would be 
> preferable, too!).
> 
> ==================
> David L. Smith

How about putting a cache memory on the board too.  (about 256K should
do the trick, I would think (:-), but seriously, at today's memory
prices, why not!?)  There are boards available that do just that with
a 65C02 running at about 6MHz (They claim a 3 to 1 speedup, anyway)
But they use high-speed static ram (about 8K, I think I remember
reading) for the cache.  And indeed, they do slow down when accessing
the video memory and the IO block between C000 and CFFF.  Still, why
not put a *big* chunk of memory at addresses above (say) 2MB, that
can be accessed fast by the 65816 and only make it slow down to
access the low order 128K that it shares with the 6502.  The logic of
putting it above 2Meg is to get it out of the way of things like
RAMWorks or Slinkey, if some way can be worked out to allow the 65816 to
access that sort of memory directly.  (Probably not, probably have to poke
soft switches just like everybody else.  But it's worth a try! I say.)
Does anybody know enough about the IIe MMU chip to say one way or
another?

I think it has possibilities!

Check the latest issue (or maybe the one before) of Call A.P.P.L.E
for an article about the speedup card.  I'll bet the people who make
that card could be talked into bringing out a 65816 version with only
a little prodding.  (Unfortunately, I forget the name of the people
who make the card, and I cant find the article right now.  I'll post
that info though, when I can find the article.)

Rick Thomas
ihnp4!attunix!rbt

ix469@sdcc6.UUCP (David L. Smith) (11/06/85)

Just in case anyone's interested...

Latest prices on 65816, single quantity, from Western Design Center:

		2Mhz - $95
		4Mhz - $104


==================
David L. Smith
UC Sandy Eggo
{ucbvax, ihnp4}!sdcsvax!sdcc6!ix469

"...here was a being to be reckoned with, a being who had hitchhiked
across the galaxy and still knew where his towel was."